A scalable, GFP-based pipeline for membrane protein overexpression screening and purification
We describe a generic, GFP-based pipeline for membrane protein overexpression and purification in Escherichia coli. We exemplify the use of the pipeline by the identification and characterization of E. coli YedZ, a new, membrane-integral flavocytochrome. The approach is scalable and suitable for high-throughput applications. The GFP-based pipeline will facilitate the characterization of the E. coli membrane proteome and serves as an important reference for the characterization of other membrane proteomes.
Membrane proteins (MPs) account for 20%–25% of all open reading frames in sequenced genomes, and fulfill a wide range of central functions in the cell (Wallin and von Heijne 1998). However, our knowledge of this important class of proteins is still poor, mainly because of a lack of generally applicable approaches to the overexpression and purification steps that precede functional and structural analysis. Novel approaches in these areas are required to facilitate and speed up MP research.
The bacterium Escherichia coli is still the most widely used vehicle for MP overexpression. Overexpression in the cytoplasmic membrane is preferred to overexpression in inclusion bodies, since the isolation of functional MPs from the membrane is usually more successful than refolding from inclusion bodies (Drew et al. 2003). Green fluorescent protein (GFP) fusions can be used to facilitate the monitoring of MP overexpression in the cytoplasmic membrane (Drew et al. 2001). If the fusion protein ends up in inclusion bodies, GFP does not fold and is therefore not fluorescent; in contrast, if the fusion is expressed in the cytoplasmic membrane, GFP folds properly and is fluorescent. GFP is only fluorescent in the cytoplasm of Escherichia coli (Drew et al. 2002), which means that GFP-based screens work only for MPs that have their C terminus located in the cytoplasm. Recently, nearly all E. coli cytoplasmic MPs were fused to GFP for a membrane proteome topology screen (Daley et al. 2005). Approximately 80% of all E. coli cytoplasmic MPs have a cytoplasmic C terminus, and thus GFP can be used to monitor the overexpression levels of the majority of E. coli MPs (Daley et al. 2005).
Here, we present a generic pipeline for rapid overexpression screening, detergent extraction, and purification of MPs based on a simple MP-GFP fusion approach. We show that milligram amounts of pure functional MP can be obtained for E. coli cytoplasmic MPs from liter-scale cultures. The approach is scalable and thus suitable for high-throughput applications.
GFP-based MP overexpression screen in E. coli
Since fluorescence is one of the most convenient ways to follow a protein overexpression and purification procedure (see Waldo et al. 1999), we sought to establish a generally applicable and easily scalable pipeline for MP overexpression and purification based on His8-tagged GFP fusions. To facilitate the removal of the GFP moiety, we also included a tobacco etch virus (TEV) protease site between the MP and GFP-His8 (see Materials and Methods).
The first step in the pipeline development was to establish a generally applicable way to rank MPs according to their overexpression levels. To this end, 48 genes coding for E. coli MPs were selected based on their Cin topology from a library of MP-GFP fusions covering almost the whole E. coli membrane proteome (Fig. 1; Daley et al. 2005). For enhanced expression, cells were cultured at 25°C after induction, and expression was tested in 1 mL and 1 L culture volumes. We did not observe significant differences in expression levels due to the different culture volumes (data not shown); i.e., 1 mL was found to be a convenient culture volume for rapid and reliable overexpression screening. Using a standard protocol, nine MP-GFP fusions were isolated from 1 L cultures (Fig. 1; see “GFP-based purification scheme,” below). There is a good correlation between GFP fluorescence measured in whole cells and the amount of MP-GFP fusion that can be purified, indicating that whole-cell fluorescence is indeed a useful indicator for the overexpression of MP-GFP fusions.
GFP-based solubilization screen
The GFP moiety in the MP-GFP fusion makes it possible to rapidly monitor the ability of different detergents to extract the overexpressed fusion protein from the membrane. Table 1 shows results for the YbaT-GFP fusion. Although the ultimate choice of detergent will depend also on the ability to preserve the MP in a fully functional state, poorly extracting detergents can be eliminated quickly in this step. GFP fluorescence is a good and time-saving alternative for the gel electrophoresis/Western blotting experiments usually used in detergent screens.
GFP-based purification scheme
To optimize the final step in the pipeline, we purified nine MP-GFP fusions (Fig. 1). These fusions differ widely in size and are, as inferred from the whole-cell GFP fluorescence levels, overexpressed to different levels. Fusion proteins were purified using a combination of IMAC and size-exclusion chromatography (see Materials and Methods). The GFP moiety of the MP-GFP fusion allows the purification to be followed visually; e.g., binding efficiency of a fusion to a column or precipitation can be seen directly. The GFP moiety of the MP-GFP fusion makes it also possible to quickly and accurately determine protein concentrations.
Recovery of functional MP-GFP fusions
The TEV protease is functional in the presence of many detergents (Mohanty et al. 2003), and we reasoned that inclusion of a TEV site between the MP and the GFP-His8 moiety should make it possible to recover intact, full-length MP from MP-GFP fusions. To test this final step in the pipeline, purified YbaT-GFP (a putative amino acid transporter), GltP-GFP (a glutamate transporter) (Wallace et al. 1990), and YedZ-GFP (a protein of unknown function) were digested with His-tagged TEV protease (Fig. 2A). The digests were almost complete and pure MP could be obtained by removal of undigested MP-GFP fusion, clipped-off GFP-His8, and His-tagged TEV protease by batch-binding to a Co-Talon resin (Fig. 2A). GFP fluorescence can be used to monitor both the effectiveness of the TEV digestion and the purity of the recovered MP (Fig. 2B).
To test whether the isolated MP is functional, purified GltP was reconstituted into lipid vesicles, and its activity was compared to purified GltP-His8 (Fig. 2C). There was no difference in the glutamate uptake activity between GltP recovered from GltP-GFP and purified GltP-His8.
YedZ attracted our attention since cells overexpressing YedZ-GFP were not green but orange, suggesting the presence of some kind of cofactor. Topology studies have shown that YedZ consists of six transmembrane segments connected by very short loops, with both the N- and C-terminal ends in the cytoplasm (Drew et al. 2002). YedZ belongs to a bacterial protein family of unknown function, UPF0191 (http://www.sanger.ac.uk), and is coded in the same operon as YedY, a periplasmic molybdoenzyme (Loschi et al. 2004). Its orange color suggests the presence of some kind of cofactor, although none of the Web-based prediction tools we used to analyze its sequence identified any potential cofactor binding motifs. To test the pipeline also on a potentially cofactor binding and previously uncharacterized protein, we decided to study YedZ further.
The recovered YedZ protein was, just as cells overexpressing YedZ-GFP, orange (Fig. 2D). Optical spectra of the purified YedZ protein were recorded under oxidizing and reducing conditions (Fig. 3A). Under reducing conditions, the optical spectrum shows a maximum at 558 nm, indicating that YedZ contains a cytochrome b. Moreover, mass spectrometry (MS) showed that YedZ contains a cofactor with a molecular weight of 617 Da, which corresponds with the mass of heme b (Fig. 3B). YedZ contains a single heme, as determined by the pyridine hemechromogen assay (Zhu et al. 1999; Barber et al. 2002).
The absorption spectra of YedZ are typical of that of a cytochrome b, except for the broad and poorly resolved peak in the 450–500 nm region, most apparent under oxidizing conditions. This peak is an indication that YedZ also has a bound flavin (Barber et al. 2002). The flavin content of YedZ was determined by reverse-phase liquid chromatography after extraction of the chromophore (Fig. 3C). YedZ contains FMN rather than FAD, with a molar ratio of 0.7 FMN per YedZ molecule. The FMN content was further confirmed by a fluorescence assay (Burch et al. 1957). The determined FMN/protein ratio of 0.7, and the fact that FMN can be extracted from YedZ, shows that the flavin is not covalently attached to the protein. The FMN was not detected in MS, most likely because of the rather harsh poros R1 micropurification, that was required to clean up the YedZ sample for MS (see Materials and Methods).
We conclude that the GFP-based pipeline is compatible with the purification of proteins containing non-covalently bound cofactors; our simple purification procedure will facilitate the further characterization of the YedZ protein, which together with the periplasmic molybdoenzyme YedY forms a novel kind of nitrate reductase (D. Drew and J.W. de Gier, unpubl.).
Overexpression of MP-GFP fusions in Lactococcus lactis
Lactococcus lactis, a promising new host for the overexpression of eukaryotic MPs, was used to explore if the GFP-based pipeline can be transferred to other systems (Kunji et al. 2003). The overexpression of eukaryotic MPs in the commonly used prokaryotic overexpression systems, such as E. coli, is notoriously difficult. It has been shown recently that the bacterium L. lactis may be an attractive alternative for the overexpression of MPs (Kunji et al. 2003). L. lactis is an easy-to-handle and low-cost system with little tendency to produce inclusion bodies. To establish if the GFP-based pipeline can be used also in L. lactis, we chose the human KDEL-receptor (KDELr) as a test case. It has been reported that 6 μg of functional KDELr can be expressed per liter of L. lactis culture (Kunji et al. 2003).
KDELr-GFP was, just as KDELr, expressed in the L. lactis membrane (Fig. 4A). Specific binding of tritium labeled YTSEHDEL peptide (a KDELr ligand) to membranes isolated from cells overexpressing KDELr-GFP was >10 times higher than for membranes isolated of cells expressing the KDEL-receptor without GFP (Fig. 4B). Based upon GFP fluorescence, good estimates of the overexpression level could be obtained even from a 200 μL culture (data not shown), which means that overexpression screening can be done in a 96-well format for L. lactis.
We conclude that GFP can be used to monitor MP-GFP overexpression in L. lactis and that GFP seems to stabilize overexpressed KDELr, resulting in the production of more functional protein. We have previously made a similar observation of increased overexpression of some GFP-tagged MPs in E. coli (Drew et al. 2003).
We have established a generic, GFP-based pipeline for rapid overexpression screening, detergent selection, and purification of functional MPs in E. coli. The use of the pipeline is exemplified by the identification and characterization of E. coli YedZ, the first membrane-integral flavocytochrome found to date. The GFP-based approach makes it possible to monitor all steps in the pipeline in a very easy manner. The approach is scalable and thus suitable for high-throughput applications. The GFP-based pipeline will facilitate the characterization of the E. coli membrane proteome and serves as an important reference for the characterization of other membrane proteomes.
Materials and methods
Expression of MP-GFP fusions in E. coli
Genes encoding MPs were amplified by conventional PCR, and cloned into a modified pET28(a+) vector that harbors the TEV protease recognition sequence (ENLYFQ↓G) followed by a C-terminally 8-histidine tagged GFP. Vectors harboring the MP-GFP fusions were transformed freshly for each experiment into BL21(DE3)pLysS, and cells were grown at 37°C on Luria broth (LB) medium containing 50 μg/mL kanamycin and 30 μg/mL chloramphenicol. We used plasmids encoding cytoplasmic resistance markers rather than periplasmic ones (e.g., β-lactamase) to avoid an extra workload for the Sectranslocon, which is involved in both the translocation of proteins across the cytoplasmic membrane and assembly of (overexpressed) MPs into the cytoplasmic membrane (Drew et al. 2003). One milliliter cultures were grown in 2-mL tubes in a thermomixer (Eppendorf) at 900 rpm, and 1 L cultures were grown in 3-L baffled conical flasks in an Innova 4330 (New Brunswick Scientific) shaker at 250 rpm. When the cultures had reached an OD600 0.4–0.5, the temperature was switched to 25°C and MP-GFP expression was induced for 4 h with 0.4 mM isopropyl-β-D-thiogalactoside (IPTG). For 1 mL cultures, cells were harvested by centrifugation at 20,100g for 3 min; 1 L cultures, at 6200g for 10 min. Cells were washed once in an equal volume of PBS buffer. Cells taken from 1 mL of culture were resuspended in 200 μL PBS and transferred to a 96-well plate, and GFP emission was measured at 510 nm, with an excitation wavelength of 485 nm, on a SpectraMax Gemini (Molecular Devices).
GFP-based detergent screen
The detergents n-dodecyl-β-D-maltopyranoside (DDM) (1% w/v), n-undecyl-β-D-maltoside (1% w/v), n-decyl-β-D-maltopyranoside (1% w/v), cymal 7 (1% w/v) or cymal 6 (2% w/v), and n-octyl-β-D-glucopyranoside (2% w/v) (all from Anatrace, Inc.) were added to membrane suspensions (3 mg of protein/mL in 1× PBS), and samples were incubated for 1 h at 4°C under mild agitation. Nonsolubilized material was removed by centrifugation at 140,000g for 1 h at 4°C, and GFP emission was subsequently measured in both the resuspended pellet (nonsolubilized material) and solubilized material as described above.
Purification of MP-GFP fusions
Cells overexpressing MP-GFP fusions from a 1 L culture were broken by means of French pressing. Membranes were isolated by means of ultracentrifugation and subsequently solubilized in PBS with 1% (w/v) DDM. The suspension contained ∼3 mg of protein/mL and was cleared by ultracentrifugation (30 min, 150,000g). The solubilisate, to which 10 mM imidazole was added, was loaded onto a newly packed 5 mL Ni-NTA superflow resin column (Qiagen) at a flowrate of 0.5 mL/min. The column was washed with PBS containing 0.1% DDM and 20 mM imidiazole for 20 CVs (column volumes). The column was subsequently washed with another 20 CVs in the same buffer containing 40 mM imidiazole. MP-GFP fusions were eluted with PBS with 0.1% DDM and 250 mM imidiazole. Eluted fractions were subsequently concentrated and loaded onto a Superdex 200 gelfiltration column (Amersham Pharmacia) in PBS with 0.1% (w/v) DDM. The amount of purified MP-GFP fusion was measured using the BCA assay (Pierce) and GFP emission as described above using a GFP standard. The two different ways of measuring the amount of purified MP-GFP fusion were consistent.
Recovery of MPs from MP-GFP fusions
One milligram of MP-GFP fusion was incubated overnight at 10°C with 1 mg of His10-TEV protease. The incubation mixture was then batch-bound to 0.5 mL Co-Talon resin (Clontech). After 1-h incubation at 4°C under mild agitation, the resin was spun down by centrifugation at 4000g for 5 min, and unbound material, i.e., the MP that had been clipped off from the MP-GFP fusion, was collected from the supernatant. Pelleted resin was washed several times, and the wash solution together with the original supernatant was passed through a filter to exclude any loose resin grains. Material bound to the resin, i.e., His10- TEV, cleaved GFP-His8, and uncleaved MP-GFP fusion, was eluted with 250 mM imidazole. Five micrograms of protein (as determined with the BCA assay, Pierce) of all fractions was assayed by SDS-PAGE/Coomassie brilliant blue R-250 staining.
Reconstitution of GltP in proteoliposomes and transport assays
Equal amounts of GltP recovered from a GltP-GFP fusion and purified GltP-His8 were reconstituted in proteoliposomes. For assays of L-glutamate uptake driven by artificial gradients, the proteoliposomes were washed twice with 20 mM morpholine-ethanosulfnic acid (Mes) (pH 6) and 100 mM potassium acetate and concentrated. Proton motive force driven uptake was initiated by diluting the proteoliposomes 75-fold into 120 mM Mes, 100 mM methylglucamine, 0.7 mM valinomycin, and 1.3 mM L-[14C]-glutamate prewarmed at 30°C. Both a proton motive force and sodium ion motive force was created by dilution into the same buffer containing 100 mM NaOH instead of methylglucamine. Control experiments were performed by diluting the proteoliposomes into the buffer with which they were loaded (Gaillard et al. 1996).
Characterization of YedZ
Recovered YedZ was oxidized in 2 mM [Fe(CN)6]3−. Subsequently, absorption spectra were measured with a light spectrophotometer (SHIMADZU, UV-1061), before and after reduction with sodium dithionite (final concentration, 10 mM).
For MALDI-TOF MS identification of cofactors, YedZ was micropurified using poros R1 (C4 like material) microcolumns (Gobom et al. 1999). A saturated solution of sinapinic acid (20 mg/mL) in 75% acetonitrile, 1% formic acid was used as an elution buffer. After deposition onto the MALDI target, the sample was washed with 10 mM diammonium citrate buffer to remove matrix adducts. The protein/heme ratio of YedZ was determined with a pyridine hemechromogen assay (Zhu et al. 1999; Barber et al. 2002).
To distinguish FMN from FAD and estimate the amount of flavin per YedZ molecule, YedZ was heated at 80°C for 45 min, and diluted in 50 mM ammonium acetate (pH 5.5). No degradation of FAD was observed during similar treatment of control samples. Samples were sonicated for 60 sec, and debris was removed by centrifugation. YedZ supernatant, FAD and FMN were adjusted to a concentration of 19.4 μM and injected into a HPLC on a C18 column with mobile phase 66%50 mM ammonium acetate pH 5.5/34%methanol. The flavin content of YedZ was also monitored using the method described by Burch (1957).
Expression of MP-GFP fusions in L. lactis
The Nisin A expression system (Kunji et al. 2003) was used to express MP-GFP fusions in L. lactis. GFP fluorescence was monitored as described above. The isolation of membrane vesicles, KDELr Western-blotting experiments, and ligand binding assays in the presence of CHAPS (0.5%) were performed as described (Kunji et al. 2003).
Research in the laboratory of J.W.d.G. is supported by the Swedish Research Council, the EMBO YIP, STINT (joint grant with K.J.v.W.), and the Marianne and Marcus Wallenberg Foundation. Research in the laboratories of E.K. and J.H. is supported by the MRC and the EMBO YIP. Research in the laboratory of P.G. is supported by the Swiss National Science Foundation. D.D. was recipient of ESF and EMBO short-term fellowships. D.J.S. was supported by a long-term fellowship of the Human Frontier Science Program Organization. M.L. is recipient of a fellowship from the Swiss National Science Foundation. D.O.D. is recipient of an EMBO long-term fellowship. John Walker and Gunnar von Heijne are gratefully acknowledged for their support.